t-test not using filtered data for two condiitions

Everything related to the development of modules in jamovi
Post Reply
jrhenkelmann
Posts: 8
Joined: Wed Aug 29, 2018 10:17 pm

t-test not using filtered data for two condiitions

Post by jrhenkelmann »

Hi there!
First of all a very big thank you for this whole project... it feels so good that finally a serious and easily usable statistics program that is in many aspects even better than expensive proprietary solutions exists!

My request is about a certain behavior I like in SPSS:
When conducting a t-test there, you can filter the data of a variable that has multiple states to only two in order to perform a t-test. Unfortunately, the t-test in jamovi does only see the complete, unfiltered data and throws an error.

Would be great if the t-test (or maybe this is a general issue effecting all modules) would include the filter before throwing an error based on the unfiltered sample.

Greetings,
Jens
User avatar
jonathon
Posts: 2613
Joined: Fri Jan 27, 2017 10:04 am

Re: t-test not using filtered data for two condiitions

Post by jonathon »

hi,

if the t-test is seeing the unfiltered data, then that is a bug. could you attach a .omv file?

with thanks

jonathon
jrhenkelmann
Posts: 8
Joined: Wed Aug 29, 2018 10:17 pm

Re: t-test not using filtered data for two condiitions

Post by jrhenkelmann »

Hi!
Thanks for your quick answer... it seems there was something buggy going on with the filter in general... while I was compiling a new dataset to send to you the filter started working as intended ... then I tried again with the original dataset and filtering did not work at all anymore... then jamovi crashed a couple of times but now it's working as intended... no idea what was going on here :)

Greetings,
Jens
jrhenkelmann
Posts: 8
Joined: Wed Aug 29, 2018 10:17 pm

Re: t-test not using filtered data for two condiitions

Post by jrhenkelmann »

In connection to this I was wondering: Would it not be a good idea to have some sort of RESET button to just go back to the plain data without any analyses, filtering etc.? I have the suspicion that somewhere in that area something went wrong...
Oh and one other feature I was wondering about: In general I like the "dynamic" analyses, i.e. that they change automatically when something is changed... yet in some cases it would be very handy if you could freeze an output, e.g.if you want to compare two analyses where you used different filters... what do you think?
User avatar
jonathon
Posts: 2613
Joined: Fri Jan 27, 2017 10:04 am

Re: t-test not using filtered data for two condiitions

Post by jonathon »

hey,

our finding has been that the filter system is pretty reliable, but it sounds like there's an issue in there i'll try and track down. i've been planning a bit audit of that body of code for a while now.

wrt freezing output, freezing itself isn't really compatible with our reproducible philosophy, but i do have some more distant plans where you can define "subsets" of the data, and have analyses using different filters side-by-side.

cheers

jonathon
jrhenkelmann
Posts: 8
Joined: Wed Aug 29, 2018 10:17 pm

Re: t-test not using filtered data for two condiitions

Post by jrhenkelmann »

Hi,
Great you're looking into that... somewhere else on the forum I saw someone reporting the "failure to connect to server" issue in connection to t-test... this happened to me too, after the filter did not work anymore... maybe this narrows it down a bit.
I'm amazed about this subset idea, that indeed sounds much better than just freezing the output! Looking forward to it!

Greetings,
Jens
Post Reply